Make properly pip installable without needing build isolation (uv pip install now works too)#402
Conversation
|
I tested this by creating an empty workspace/venv where i pip installed I also verified that I could run the benchmark. Please give this a spin and see if it works for you. |
|
If Dao-AILab/causal-conv1d#26 is merged, then:
Should no longer be true. |
… install now works too)
|
@albertfgu @tridao any chance this can be accepted (i'll update it if so), otherwise this can be closed. |
|
@albertfgu @tridao Hi, Can you approve this pr? btw @kszlim thanks, good work! |
|
Please approve this pr. It also fixed my installation issue. |
|
Hi @tridao when is a new release to pypi expected with this fix? The installation of the latest available package fails. |
… install now works too) (state-spaces#402) Co-authored-by: Tri Dao <tridao@users.noreply.github.com>
Now you can just pip install mamba_ssm and it will work, previously it would throw unless you had
packagingandtorchalready installed. It would also always fail if you had build isolation on (which will be on by default in the future withinpip).As a side effect, now
uvcanuv pip install mamba_ssmproperly now.Take note that installing
causal-conv1dvia the optional extra will still fail with build isolation due to that package needing similar changes to this one.